A practical reason the vision for AI adoption seems unserious is the collective overestimation of our processes.
You can tell an agent “Put X information into Y format and update Z tracker” but only if the information is documented, the format is consistent, and the tracker is up to date.